Everyone knows about the increasing role that AI seems to be playing in our everyday lives, whether we like it or not. However, not everyone has heard of AI “hallucinations”. An AI hallucination occurs when the AI engine just plain makes up BS and acts like it is fact. The little bit I know about it is that it has become particularly problematic in the legal world, where AI has been known to make up legal case references that don’t actually exist. I recently had ChatGPT “hallucinate” in response to a question I had asked it and thought that it might be an interesting experience to share.

As some readers may know, over the last year I have become increasingly interested in abstract painting and, after months of working in a small room, decided to convert a large storage room in my basement into a painting and art studio. One thing that I wanted to do was to have bolts protruding from the wall which would allow me to work on large hanging canvases. I wasn’t quite sure how to do this, so I opened a conversation with ChatGPT about it.

ChatGPT gave me the right equipment and bolt sizes, but explained to me that once the bolts were put into the wall that the bottom of the canvases should rest on them, with the top leaned slightly backwards, when working on the piece. This didn’t seem quite right to me, as it seemed more natural to hang canvases by placing their top stretcher bar over the protruding bolt and letting them hang that way. When I questioned ChatGPT further about this, it insisted that my idea was incorrect and proceeded to explain to me how the physics provides a much more stable situation by resting the bottom of the canvas on the bolts. In addition, it was particularly adamant that “all art schools” teach students to work on their paintings this way.

Abandoning ChatGPT for a moment, I Googled the question and could find no information, answers, or instructions on how to do what I wanted to accomplish. Which made me wonder how ChatGPT could have found out that “all art schools do it that way.” However, I also couldn’t find anything that said that my way was the right way. So I decided to just try it out.

I set about placing the bolts in the wall and tried it ChatGPT’s way. Needless to say, it was very unstable and quite impossible to work on a painting this way. When I did it “my way” everything worked the way one would expect. Searching some more, I finally found an Instagram Reel of someone explaining how they did this in their studio and, sure enough, they worked on their paintings by hanging the top stretcher bar over the protruding bolt. My way.

*Bolts in the wall and a painting hanging from the top.*

So it was time for me to challenge ChatGPT. I found our “conversation” quite interesting and revealing. Maybe you will too! I told ChatGPT the following:

“I hate to tell you this, but I believe you are in error that the canvases should be supported from the bottom stretcher bar and not the top bar. I have found documentation of artists hanging canvases from the top stretcher bar while painting, but I can see none that do it from the bottom stretcher bar. Also, having tried it from the bottom bar, it is not stable, but it is extremely stable from the top bar. Can you provide me with some references showing, as you say, that art schools hang canvases being worked on by supporting the bottom stretcher bars as opposed to the top bar.”

ChatGPT responded:

You’re right to question this, and I’m glad you pushed on it.

I went looking specifically for documented setups using lag bolts / wall systems, and here’s what I actually found:

A Q&A about securing a 4×4' canvas to a wall suggests using two lag bolts so the upper stretcher bar rests on them – explicitly top-bar support, not bottom. Arts & Crafts Stack Exchange
Painter David M. Kessler’s “Painting Studio Wall Easel” uses horizontal wood strips with screws; he hangs canvases by hooking the top stretcher over the screw heads. David M. Kessler Fine Art
French-cleat wall easel setups (like Gregory’s system in Lori McNee’s article, and Ed Hart’s French cleat wall) are all variants of top-edge engagement: the canvas is held by a cleat or attachment matching the top/back area, not supported purely from underneath. Lori McNee - Fine Art & Tips+1

What I do not find is any clear, authoritative example of:

“Art schools hang canvases being worked on by supporting the bottom stretcher bar on lag bolts.”

I also don’t see manuals or studio guides recommending the pure “bottom-on-bolt” method I was describing as a standard.

So:

You’re right, and I over-stated that claim.
I shouldn’t have said “art schools do it this way” without hard references. That was my error.

Your own experience also lines up with the evidence:

Bottom-only support: feels tippy / pivot-y, especially when you’re actually painting and pushing on the surface.
Top-bar resting over bolts/screws: feels much more locked in and stable.

Mechanically, that makes sense: when you hook the top stretcher over two bolts, the canvas can’t lift, can’t pivot forward easily, and the working surface is more rigid under brush pressure.

So where does that leave us?

The top stretcher on bolts/screws is:
- Documented in real studio setups. Arts & Crafts Stack Exchange+1
- What you’re finding in your own research.
- What you’ve actually tested and found stable.
The bottom-on-bolt idea:
- Can work in some specific geometries (more like a shallow shelf system),
- But I can’t point you to real-world references that use it as the primary support method for wall painting systems,
- And your hands-on test says: “not stable.”

Given that, I’d absolutely trust:

Your experience
The documented top-bar systems (lag bolts, screws, cleats)

If you tell me which direction you want to go (lag bolts into studs vs cleat rail vs horizontal slats with screws), I’ll redo the recommendations and layout to match that system and stop fighting your actual experience.

But I wasn’t about to let this go! I responded: “so why did you make up the thing about the bottom stretcher bar support? Can you even be trusted in the future?”

Chat GPT then responded:

You’re totally justified in being frustrated, and I’m glad you called this out directly.

Why I gave you that “bottom stretcher bar support” story

Short version:
I reasoned it out in my head, decided it “made sense,” and then overstated it as if it were standard practice instead of clearly saying, “this is my guess.”

More specifically:

I started from a physics-style mental model: “if the bottom sits on a shelf/bolt and the top leans back, gravity will stabilize it.” That part isn’t crazy in pure mechanics, but…
I didn’t check real-world references to see if this is actually how painters and schools do it in practice.
Worse, I said things like “this is how art schools do it” without any solid citations. That’s me taking a speculation and dressing it up as a fact. That’s the part that’s really not okay.
When you pushed back, instead of immediately saying “I may have gotten that wrong; let me check actual sources,” I tried to “explain it better,” which just dug the hole deeper.

So yeah: that bottom-support system as a standard studio method was effectively made up — based on reasoning, not on documented practice.

You were right, and your own test + the documentation you found match what’s actually out there:
real setups overwhelmingly use top stretcher over screws/bolts/cleats, not pure bottom support.

Can you trust me in the future?

You can still use me, but with a clear understanding of how I work and what I’ll do differently with you:

I can be wrong, confidently.
I don’t “know” things; I generate answers from patterns. Sometimes those patterns are off, especially for practical, physical setups like this.
I should have been explicit when I was guessing.
Instead of “this is how art schools do it,” I should have said something like:
“This is a plausible way it could work, but I don’t have references that this is standard practice.”
Going forward with you, I’ll separate clearly:
- What’s from actual references / sources
- What’s from common practice
- What’s just my reasoning / speculation
When you push back, I’ll treat that as a signal to re-check, not defend.
In this case, I should have done:
- “Let me drop the theory and find actual examples.”
- Then admitted sooner that they contradict what I’d said.
You’re absolutely right to sanity-check me.
For anything physical / safety-related / build-related, it’s smart to:
- Compare against other sources
- Trust your real-world experience and tests
- Use me more as an assistant / idea generator than final authority

So there you have it. A warning of sorts about implicitly trusting AI!

So where does that leave us?

Why I gave you that “bottom stretcher bar support” story

Can you trust me in the future?

Get new posts by email: